Table 1 Summary of included studies.

From: Health benefits and harms of mammography screening in older women (75+ years)—a systematic review

Study Reference

Country

Participant population/age of women included

Study design

Screening context/setting

Screening process/screening interval

Comparison made

Total numbers in comparison groups

Follow-up/collection period

Risk of Biasa

Advani [16]

USA

66–94 years

Cohort study

Women who underwent screening mammography in 1999–2010 in a Medicare-linked BCSC database. Outcomes ascertained via linkage with regional SEER programmes or state tumour registries

At least one screening mammogram

66–74 vs 75–84 vs 85–94

304,334 vs 190,180 vs 30,346

Follow-up: Within one year or until next screen

Collection period: 1999–2010

Serious

Arleo [17]

USA

Using Cancer Intervention and Surveillance Modelling Network (CISNET) to develop models

Modelling study

Using CISNET models to simulate and compare three screening mammography recommendations

Annual and biennial 75–79 years and 75–84 years

Different screening recommendations

1000 women screened

N/A

Low

Bennett [18]

England

All women self-referred to NHS screening programme 71 years and older

Screening evaluation study

Analysed data for all women in England over the age of 70 who self-referred in the 2-year period 2005–2008.

All women self-referred to NHS screening programme and attended one screen during the collection period

70–74 vs 75 and older

71–74: 86,743, 75+: 53,167

No follow-up

Collection period:3 years (2005–2008)

Serious

Boer [19]

The Netherlands

MISCAN (Micro Simulation Screening Analysis) model starts with women aged 40, but paper focuses on 70 years and older

Modelling study

The MISCAN model (simulated model with 2-yearly screening carried out during a period of 27 years)

2-year interval

Optimistic variant vs pessimistic variant and 69 vs 99

Assuming 100% attendance rate; and realistic attendance and invitation starting at 51 years

N/A

High

Braithwaite [21]

USA

66–89 years

Cohort study

Data linkage, registries, women receiving (free) mammograms between 1999 and 2006

Annual (9–18 months) or biennial (>18–30 months)

66–74 vs 75–89

2993 women aged ≥66 and with breast cancer and 137949 women aged ≥66 and without breast cancer

Follow Up: Not stated, max of 7 years (1999–2006)

Collection period: 1999–2006

Serious

Braithwaite [20]

USA

66–99 years

Cross-sectional study

Not reported

At least one screening mammogram

66–74 vs 75–84 vs 85-99

6587 (1.2%) were followed by biopsy within 90 days among 537254 screens (171,636 women).

No follow-up Collection period:1999–2010

Critical

Cate [22]

USA

75 years and older

Cross-sectional descriptive

Screening mammography

Not reported

No comparison

2057 in total

No follow-up

Collection period:2013–2014

Critical

Demb [23]

USA

66–94 years

Cohort study

No less than 1 screening mammogram between 66 and 94 years

All women self-referred to screening NHS programme

66–74 vs 75–84 vs 85–94

222,088 in total

Follow-up: Median of 107 months (IQR = 65–120);

Collection period: 1999–2010

Serious

Destounis [24]

USA

75 years and older

Cross-sectional descriptive

Data from 763,256 screening mammograms at Elizabeth Wende Breast Care (all ages)

Mammography but no further details.

No comparison

76,885 screening participants aged 75+

No follow-up

Collection period: 10 years 2007–2017

Critical

El-Zaemey [4]

Australia

65 years and older

Cohort study

Free screening for all women ≥40 years offered every 2 years

All women aged up to 75 years are invited to participate in biennial screening programme. Women age 75years and older are self-referred.

65–69 vs 70–74 vs 75 and older

39,886 vs 26,432 vs 8763 (number of screens)

Follow-up: Up to 3 years

Collection period: 2015–2017

Serious

Erbas [25]

Australia

40 years and older

Cross-sectional study

Free screening for all women ≥40 years offered every 2 years

A second or subsequent screen in a woman with at least one prior negative screen within the biennial programme.

40–49 vs 50–69 vs 70–74 vs 75 and older

27,661 vs 506,117 vs 87,545 vs 24,699 (646,022 in total)

No follow-up

Collection period: 1993–2000

Serious

Garcia-Albaniz 2020

USA

70–84 years

Cohort study

20% of Medicare fee for service beneficiaries. No history of BC, continuous enrolment in Medicare Parts A and B for 12+ months

Stop screening (women who do not have any screening after baseline), Continue screening (women who continue annual screening (with 3 mth grace period)

Continue vs stop screening

1,403,735 (aged 75–84)

Follow-up: Median of 16 months (IQR = 14–32); Collection period: 1999–2008

Serious

Hartman [27]

USA

75 years and older

Cross-sectional descriptive

All screening mammograms performed at institution from 2007 to 2013, outcome is screen detected cancers at institution, free screening for all women annually

Annual breast cancer screening

No comparison

4424 in total

No follow-up

Collection period: 2 February 2007–31 December 2013

Critical

Jansen [28]

The Netherlands

A computer model for the simulation of breast cancer screening (MBS) to calculate results of screening in terms of lifetime ≥20 years

Modelling study

The MBS model (simulated model with a stable Swedish population of one million women)

Annual mammography with single view

No comparison

One million

Lifetime (simulated)

Moderate

Jansen [29]

Sweden

A computer model for the simulation of breast cancer screening (MBS) to calculate results of screening in terms of lifetime ≥20 years

Modelling study

Using Swedish two county study and the MBS model

Annual mammography with single view

No comparison

One million

Lifetime (simulated)

Moderate

Kregting [30]

The Netherlands

40-84 years

Modelling study

920 breast screening strategies with varying starting ages (40–60) and stopping ages (64–84)

Screening intervals of 1-4 years

Many - have chosen annual screening 40-75 to annual screening 40-84

N/A

N/A

Moderate

Lansdorp-Vogelaar [31]

USA

US cohorts 66–90 years in 2010 with average health or one of four comorbidity levels - none, mild, moderate, or severe. Using models from CISNET

Modelling study

Simulated data. Assuming all undergo regular screening starting at 50 with biennial mammography and follow individuals for their remaining lifetime.

MISCAN-Fadia (Microsimulation screening analysis – Fatal diameter) model and a G-E (Georgetown-Einstein) model

Regular screening starting at age 50 with biennial mammography

74 vs 76

Not reported

Lifetime (simulated)

Low

Malmgren [32]

USA

All women with primary breast cancer aged 50+ (75–94 years)

Cohort study

Cancer detected between 1990 and 2008, presenting at clinic. All women either patient, physician or mammography detected

Mammography but no further details

Mammography detected v physician detected v patient detected

5595 in total, 75+ = 950

Follow-up: not stated

Collection period:1990–2008

Serious

Malmgren [33]

USA

All women with primary breast cancer aged 50+ (75–94 years)

Cohort study

Cancer detected between 1990–2011, presenting at clinic (using registry database). All women either patient, physician or mammography detected

Mammography but no further details

Mammography detected v physician detected v patient detected

1162

Follow-up: Mean 7.9 years (range 1.8–21 years)

Collection period: 1990–2011

Serious

Mandelblatt [34]

USA

Using Cancer Intervention and Surveillance Modelling Network (CISNET) to simulate six models for women born in 1960 beginning at 25 yrs old

Modelling study

Using CISNET models to estimate the benefits and harms of alternate screening strategies

10 strategies each evaluated using annual and biennial schedule—a total of 20 strategies

Stopping screening

1000 women screened

N/A

Moderate

McCarthy [35]

USA

All women in the SEER database wo received a diagnosis of incident breast cancer in 1995 or 1996 and were linked with Medicare data. Aged 69 years and older

Cohort study

Cancer detected between 1987 and 1993 linked to Medicare records which indicated pattern of screening

Non-user (none in past 2 years), single user (once in the last 2 years), regular user (at least 2 screening mammograms at least 11 months apart in past 2 years)

i) Age at diagnosis

ii) Non-users v regular users

67–74 n = 4609

75–84 n = 4072

85+ n = 1086

Screen-usage

67-74: Reg 29%; Non 18%

75–84: Reg 23%; Non 21%

85+: Reg 10%; Non 33%

Follow-up: 1–8 years

Collection period: 1987–1993

Serious

McPherson [36]

USA

Caucasian women aged 65–101 diagnosed with invasive breast cancer from 1986 to 1994

Cohort study

Cancer detected between 1986 and 1994 identified from records in the Upper Midwest Oncology Registry System. Method of detection extracted

Mammography but no further details

Mammography detected v clinically detected cancers

75–79: 599 v 546; 80–84: 473 v 313; 85+: 451 v 148

Follow-up: not stated, but reports 8–9 years post diagnosis data

Collection period: 1986–1994

Serious

Park [37]

USA

40–85 years at baseline

Cohort study

Women enroled in the California Teachers study (1995–1996) followed up with data linkage up until Dec 31 2015

Routine screening

Never/less frequent v biennial v annual screening

8024 women aged 75-85 at baseline

Follow-up: 20 years

Collection period: 1995–2015

Serious

Randolph [38]

USA

All women in the SEER database wo received a diagnosis of incident breast cancer in 1995 or 1996 and were linked with Medicare data. Aged 69 years and older

Cross-sectional study

Cancer detected between 1995 and 1996 linked to Medicare records which indicated pattern of screening

Nonuser (no screening in past 2 years), single user (once in the last 2 years), regular user (at least 2 screening mammograms at least 11 months apart in past 2 years)

65–74 v 75 and older

Total 11,039.

75+ = 6813

65–74 = 4226

No follow-up

Collection period: 1995–1996

Serious

Richman [50]

USA

Women in the SEER database who had not received a diagnosis of breast cancer before 2002, but had had a screening mammogram in 2002 and were linked with Medicare data. Aged 70 years and older

Cohort study

Population setting, screening mammography and breast cancer diagnosis recorded via population data sets (Medicare and SEER database)

Screening mammograms identified in Medicare claims. Initial screen in 2002, subsequent screen in the 3 year period after 2002 screen

Subsequent screen v no subsequent screen (75–84 years)

23,613 screened, 5707 unscreened

Follow-up: Median 10 years (IQR 5.8–13.9 years)

Collection period: 2002–2017

Moderate

Schousboe [39]

USA

65–90 years

Modelling study

Markov modelling using SEER data from 2011 to 2016

Biennial

Stopping screening at 75 compared to 80, 85 and 90 - by CCS, biennial mammography

N/A

N/A

Moderate

Simon [40]

USA

Women’s Health Initiative study—75 years and older

Cohort study

Women in WHI diagnosed with breast cancer

Mammography but no further details

Mammography interval - <2 years, 2–5 years, 5+ years or none

8663 in total - not stated for 75+

Follow-up: average of 12.2 years

Collection period: Not stated (WHI study)

Serious

Simon [41]

USA

Women’s Health Imitative study—aged 50–79, post-menopausal, (75 years and older)

Cohort study

Women diagnosed with invasive breast cancer at 75 years and older. Followed up until last document contact, death or September 2010 which ever came first

Clinical trials - Annual or biannual, Observational: at the discretion of the individual

Mammography interval - <2years, 2–5 years, 5+ years or none

Total 1914: <2 years = 1398, 2-5 years = 372 5+ years = 144

Follow-up: mean 4.4 years, max 15.3 years

Collection period: Recruitment 1993–1998, Follow-up to 2010

Serious

Smith-Bindman [42]

USA

All women with Medicare

Cohort study

Linked data from Medicare screening services and State cancer registry

Any screening within 1992 and 1993

No screening

Total N = 201,537 (75–79 years), screened n = 201,537 (40%)

Follow-up: 12-24 months

Collection period: 1992–1993

Serious

Upneja [43]

USA

67–74 years

Cohort study

SEER database— women who were screened during 2015

One screening mammogram in 2015

67–74 vs 75 and older (all undergoing 2D mammography)

364,254 vs 230,685

Follow-up: 4 months

Collection period: 2015–2016

Serious

VanDijck [44]

The Netherlands

Women invited to screening at least twice, with most recent invitation having occurred over 65 years old

Case control study

Population based screening programme

Women invited to biennial single-view mammography screening. At least 2 mammography screening invitations one of which occurred when the woman is aged 65 years or older

Attended last screen or not

33 cases, 165 referents - from 65 years older - Based on Table 2: 12 cases (and 12*5 = 60 referents)

Follow-up: NA

Collection period: cases had died of breast cancer between 1977 and 1988.

Serious

VanDijck [45]

The Netherlands

Nijmegen programme: women invited to screening at least twice, with most recent invitation having occurred over 65 years old

Cohort study

Primary breast cancer patients diagnosed before Dec 1994 aged 50+

Women invited to biennial single-view mammography screening. Women aged 70 years and older invited from 9th screening round onwards

First screen, repeat screen, interval CA or non-participant

4253 invited to first screen and 33,949 invited to subsequent screens

Follow-up: 2 years (for older ages)

Collection period: 1975–1994

Serious

VanDijck [46]

The Netherlands

Nijmegen programme: women invited to screening at least twice, with most recent invitation having occurred over 65 years old. Cases died of breast cancer before Jan 1, 1994

Case control study

Population based screening programme,

Women invited to biennial single-view mammography screening. Women aged 70 years and older invited from 9th screening round onwards

No screening (no participation in index round and 4 preceding rounds)

Regular screening (participation in index round and negative mammogram in preceding round)

Otherwise (not meeting the criteria for unscreened or screened)

Aged 64+: 82 cases, 410 controls

aged 75+: 35 cases, 175 controls

- no screening 20/97

- regular screening 3/14

- otherwise 12/64

Follow-up: NA

Collection period: 1975-1994

Serious

vanRavesteyn [46]

USA

Simulation of a cohort of American women born in 1960, starting screening 50

Modelling study

Simulation

All women received biennial screening starting at 50, with cessation ages varying from 74 up to 96 years

Stopping screening at 74 v continuing screening

N/A as this is statistical modelling

N/A

Low

Vyas [48]

USA

39,006 women aged 70+ with incident BC from 2005 to 2009

Retrospective cohort study

From the Surveillance, Epidemiology and End Results (SEER) Medicare dataset. Programme collects information on newly diagnosed cancer cases from 18 population-based tumour registry’s and covers approx. 26% of the US population.

Mammography screening persistence during the 5 years period before BC diagnosis was to be determined, women who were continuously enroled in Medicare parts A/B for at least 60 months before BC diagnosis, and who were not enroled in health maintenance organisations. Persistent users represented a population who have had annual to biennial mammography screening before BC diagnosis

70–74 v 75–79 v 80 and older

All:

70–74: 12163, 75–79: 11,182, 80+: 15,661.

Persistent:

70–74: 6504 75–79: 5672, 80+: 5732

Non-Persistent:

70–74: 3270 75–79: 3019, 80+: 3933

Non-users:

70–74: 2389 75–79: 2491, 80+: 5996

Follow-up: up to 5 years (at least 60 months)

Collection period: Incident BC between 2005–2009, and 5 years prior for screening

Serious

Yang [49]

USA

Population data women aged 40 years and older

Cohort study

Pre-1977 historical cohort as America initiated screening in 1977. BC diagnosed in 1999 represents the screening cohort— using SEER data

Mammography according to the ACS guidelines at the time for the age group. Population screening in USA— data from SEER data.

1973–1976 cohort vs 1999 cohort (No screening vs screening)

Total N: 10,538 5708 vs 4830

Follow-up: 15 years for survival

Collection period: year of diagnosis: 1973–1999

Moderate

  1. aTo assess risk of bias, we used the Risk of Bias in Non-randomised studies of interventions (ROBINS-I) tool or the Quality of primary diagnostic accuracy studies (QUADAS-2) tool for observational studies and adapted the Risk of Bias tool developed by Carter et al. [15] for modelling studies using two of their criteria (i. transparent assumptions and v. data validation).