Total fish consumption and all-cause mortality or survival. Effect modification.

  • Level of consumption: Effects probably differ between different levels of intake. For this analysis, four different levels of consumption were considered, relative to nonconsumption: < 1, 1, 2, and = 3 servings/week.
    17 cohorts with 29,006 cases could be included (Table 1) after excluding the debatable results from Hirayama T [7]. Average RR's were 0.98, 0.94, 0.94, and 0.91 for < 1, 1, 2, and ≥ 3 servings/week, respectively.
  • Gender: Analysis stratified by sex can be found in table 2. Effect sizes were similar for men (13 cohorts: RR = 0.90) and women (10 cohorts: RR = 0.88). One other author did not describe data stratified by gender, but stated that effects were similar for men and women (Yamagishi K [29]).
  • Geographic area: Protective effects were stronger among US cohorts (8 cohorts: RR = 0.88) and Asian cohorts (4 cohorts: RR = 0.89) than among European cohorts (17 cohorts: RR = 0.98). Tables 3-5.
  • Other possible effect modifiers: No effect modification was found by age (Gillum RF [15]). income or BMI (Tomasallo C [31]).

Table 1. Prospective studies of total fish and all-cause mortality.
Stratified by level of consumption:
AuthorCohort nameCases< 1 serving/week1 serving/week2 servings/week≥ 3 servings/week
32) Manger MS (2010)The WENBIT137HR = 1HR = 1HR = 1HR = 0.94
31) Tamasallo C (2010)No cohort name defined177 captains, and

128 referents
Captains: HR = 1.07

Referents: HR = 0.88
Captains: HR = 1.00

Referents: HR = 0.60
Captains: HR = 1.00

Referents: HR = 0.60
Captains: HR = 1.00

Referents: HR = 0.60
29) Yamagishi K (2008The JACC Study7,008HR = 1HR = 1HR = 0.98HR = 0.95
27) Ness AR (2005)The Boyd Orr Cohort1,010RR = 0.96RR = 0.92RR = 0.98RR = 0.98
26) Nakamura Y (2005)The NIPPON DATA801,745RR = 1RR = 0.89RR = 0.88RR = 0.89
25) Folsom AR (2004)The Iowa women's Health Study4,653RR = 1RR = 0.97RR = 0.93RR = 0.93
22) Erkkilä AT (2003)The EUROASPIRE Study34RR = 0.50RR = 0.50RR = 0.50RR = 0.44
21) Osler M (2003)No cohort name defined1,329HR = 1*HR = 0.95HR = 1.14HR = 1.20
20) Barzi F (2003)The GISSI-Preventione Trial1,660OR = 1OR = 0.87OR = 0.81OR = 0.76
19) Hu FB (2003)The Nurses' Health Study486RR = 0.88RR = 0.67RR = 0.69RR = 0.59
18) Nagata C (2002)The Takayama Study1,163 men, and

899 women
Men: HR = 1

Women: HR = 1
Men: HR = 1

Women: HR = 1
Men: HR = 1

Women: HR = 1
Men: HR = 0.93

Women: HR = 0.94
17) Yuan JM (2001)The Shanghai Cohort Study2,134RR = 1RR = 1RR = 1RR = 0.84
15) Gillum RF (2000)The NHANES I Study1,236 white men,

277 black men,

1,103 white women, and

285 black women
White men: RR = 0.88

Black men: RR = 1.01

White women: RR = 1.02

Black women: RR = 0.77
White men: RR = 0.76

Black men: RR = 1.05

White women: RR = 1.02

Black women: RR = 0.79
White men: RR = 0.85

Black men: RR = 1.11

White women: RR = 0.90

Black women: RR = 0.82
White men: RR = 0.85

Black men: RR = 1.11

White women: RR = 0.90

Black women: RR = 0.82
13) Whiteman D (1999)The OXCHECK Study477RR = 1RR = 1.16RR = 1.16RR = 1.11
12) Albert CM (1998)The Physician's Health Study1,652RR = 0.90RR = 0.71RR = 0.70RR = 0.72
11) Mann JI (1997)The Oxford Vegetarian Study389DRR = 97DRR = 96DRR = 96DRR = 96
7) Hirayama TNo cohor name defined55,523 men, and

23,544 women
Men: RR = 0.81*

Women: RR = 0.92*
Men: RR = 0.78

Women: RR = 0.87
Men: RR = 0.78

Women: RR = 0.87
Men: RR = 0.76

Women: RR = 0.83
2) Daviglus ML (1997)The Chicago Western Electric Study1,042RR = 1.02RR = 1.02RR = 0.98RR = 0.85
Total number of cases: 84,529Average RR = 0.90Average RR = 0.86Average RR = 0.86Average RR = 0.83
Excluding data from Hirayama T [7]:Total number of cases: 29,006Average RR = 0.98Average RR = 0.94Average RR = 0.94Average RR = 0.91
* = Lowest level of consumption was not the reference group. RR for the lowest level of consumption was set to 1. Other RR's were adjusted for this difference in setting.


Table 2. Prospective studies of total fish and all-cause mortality.
Stratified by gender:
AuthorCohort nameCasesRelative Risk among menRelative Risk among women
29) Iso H (2007)The JACC Study9,534 men

6,444 women
HR = 0.94 (0.89-0.99; P = < 0.05)HR = 0.94 (0.88-1.00; P = < 0.05)
28) Iestra J (2006)The HALENot definedHR = 0.79 (0.59-1.06)HR = 1.04 (0.57-1.89)
25) Folsom AR (2004)The Iowa Women's Health Study4,653 women-RR = 0.93 (0.83-1.05; P = 0.15)
21) Osler M (2003)No cohort name defined826 men

503 women
HR = 1.01 (0.80-1.28; P = 0.08) for the highest vs third quartile of consumption (reference group: HR = 0.80)HR = 1.21 (0.90-1.63; P = 0.06) for the highest vs third quartile of consumption (reference group: HR = 1.03)
19) Hu FB (2002)The Nurses' Health Study4,121 women-RR = 0.68 (0.57-0.82; P = < 0.001)
18) Nagata C (2002)The Takayama Study1,163 men

899 women
HR = 0.94 (0.78-1.12; P = 0.50)HR = 0.86 (0.70-1.05; P = 0.17)
17) Yuan JM (2001)The Shanghai Cohort Study2,134 menRR = 0.79 (0.69-0.91; P = 0.01)-
15) Gillum RF (2000)The NHANES I Study1,236 white men

277 black men

1,103 white women

285 black women
White men: RR = 0.85 (0.68-1.06)

Black men: RR = 1.11 (0.68-1.81)
White women: RR = 0.90 (0.71-1.15)

Black women: RR = 0.82 (0.52-1.28)
12) Albert CM (1998)The Physician's Health Study1,652 menRR = 0.73 (0.55-0.96; P = 0.45)-
10) Salonen JT (1995)The KIHD Study78 menRR = 1.10 (0.69-1.74; P = 0.69)-
8) Dolecek TA (1992)The MRFIT522 menRR = 0.75 (P = < 0.10)-
7) Hirayama T (1990)No cohort name defined31,979 men

23,544 women
RR = 1.35 (1.20-1.52). for low vs high consumption.RR = 1.25 (1.13-1.39) for low vs high consumption.
5) Nube M (1987)No cohort name defined742 men

448 women
A protective effectAn increased risk
4) Lapidus L (1986)No cohort name defined75 women-No significant association
3) Vollset SE (1985)No cohort name defined2,587 menNo significant association (P = 0.44)
2) Daviglus ML (1997)The Chicago Western Electric Study1,042 menRR = 0.85 (0.64-1.10; P = 0.18)-
1) Fraser GE (1997)The Adventist Health Study451 men

936 women
HR = 0.89 (0.58-1.38)HR = 0.99 (0.73-1.33)
Total number of cases: 50,894 men
42,488 women
Average RR = 0.80Average RR = 0.83
Excluding data from Hirayama T [7]:Total number of cases: 18,915 men
18,944 women
Average RR = 0.90Average RR = 0.88


Table 3. Prospective studies of total fish and all-cause mortality among US cohorts:
AuthorCohort nameCasesRelative Risk (RR)
31) Tomasallo C (2010)No cohort name defined177 captains, and

128 referents
Captains: HR = 1.00 (0.63-1.58).

Referents: HR = 0.60 (0.38-0.95; P = < 0.05).
25) Folsom AR (2004)The Iowa Women's Health Study4,653RR = 0.93 (0.83-1.05; P = 0.15)
19) Hu FB (2002)The Nurses' Health Study4,121RR = 0.68 (0.57-0.82; P = < 0.001)
15) Gillum RF (2000)The NHANES I1,236 white men,

277 black men,

1,103 white women, and

285 black women
White men: RR = 0.85 (0.68-1.06).

Black men: RR = 1.11 (0.68-1.81).

White women: RR = 0.90 (0.71-1.15).

Black women: RR = 0.82 (0.52-1.28).
12) Albert CM (1998)The Physician's Health Study1,652RR = 0.73 (0.55-0.96; P = 0.45)
8) Dolecek TA (1992)The MRFIT522RR = 0.75 (P = < 0.10)
2) Daviglus ML (1997)The Chicago Western Electric Study1,042RR = 0.85 (0.64-1.10; P = 0.18)
1) Kahn HA (1984)The Adventist Health Study5,971OR = 1.04
Total number of cases: 21,167Average RR = 0.88


Table 4. Prospective studies of total fish and all-cause mortality among European cohorts:
AuthorCohort nameCasesRelative Risk (RR)
33) Tognon G (2010)The PPSW630HR = 0.96 (0.82-1.13)
32) Manger MS (2010)The WENBIT137HR = 0.95 (0.58-1.55; P = 0.98)
30) González S (2008)No cohort name defined83RR = 1.14 (0.85-1.53; P = 0.38)
27) Ness AR (2005)The Boyd Orr Cohort1,010RR = 0.98 (0.79-1.20; P = 0.8)
24) Trichopoulou A (2005)The EPIC-elderly Study (subjects without CHD)4,047MR = 1.00 (0.97-1.04)
24) Trichopoulou A (2005)The Greek part of the EPIC Study (subjects with CHD)131MR = 0.97 (0.81-1.15)
22) Erkkilä AT (2003)The EUROASPIRE Study34RR = 0.37 (0.14-1.00; P = 0.059)
21) Osler M (2003)No cohort name defined1,329HR = 1.06 (0.88-1.28; P = 0.02) for the highest vs third quartile of consumption
(reference group: HR = 0.88)
20) Barzi F (2003)The GISSI-Preventione Trial1,660OR = 0.76 (0.62-0.94; P = 0.0003)
16) Fortes C (2000)No cohort name defined53RR = 0.89 (0.46-1.74)
13) Whiteman D (1999)The OXCHECK Study477RR = 1.06 (0.71-1.59)
11) Mann JI (1997)The Oxford Vegetarian Study389DRR = 96 (76-121)
10) Salonen JT (1995)The KIHD Study78RR = 1.10 (0.69-1.74; P = 0.69)
9) Kromhout D (1995)No cohort name defined187RR = 0.96 (0.72-1.30)
5) Nube M (1987)No cohort name defined742 men, and

448 women
Men: A protective effect.

Women: An increased risk.
4) Lapidus L (1986)No cohort name defined75No significant association
3) Vollset SE (1985)No cohort name defined2,587No significant association (P = 0.44)
Total number of cases: 14,097Average RR = 0.98


Table 5. Prospective studies of total fish and all-cause mortality among Asian cohorts:
AuthorCohort nameCasesRelative Risk (RR)
29) Yamagishi K (2008)The JACC Study7,008HR = 0.92 (0.85-1.00; P = 0.08)
26) Nakamura Y (2005)The NIPPON DATA801,745RR = 1.00 (0.77-1.29; P = 0.98) for the highest vs second quintile of consumption
(reference group: RR = 1.12)
18) Nagata C (2002)The Takayama Study1,163 men, and

899 women
Men: HR = 0.94 (0.78-1.12; P = 0.50).

Women: HR = 0.86 (0.70-1.05; P = 0.17).
17) Yuan JM (2001)The Shanghai Cohort Study2,134RR = 0.79 (0.61-0.91; P = 0.01)
7) Hirayama T (1990)No cohort name defined31,979 men, and

23,544 women
Men: RR = 1.35 (1.20-1.52) for low vs high consumption.

Women: RR = 1.26 (1.13-1.39) for low vs high consumption.
Total number of cases: 68,472Average RR = 0.79
Excluding data from Hirayama T [7]:Total number of cases: 12,949Average RR = 0.89